[x265] [PATCH 03 of 13] stats: add frame statistic for average WPP benefit

[Steve Borho]

# HG changeset patch
# User Steve Borho <steve at borho.org>
# Date 1422467508 21600
#      Wed Jan 28 11:51:48 2015 -0600
# Node ID be33891b84573dc337d17fb82084aad523325fd7
# Parent  e8367f5cd43fef208eee6cf4c03974d330253f4a
stats: add frame statistic for average WPP benefit

Show how many worker threads, on average, were working on each frame. Also move
the performance statistics together at the end of the CSV line in preparation
for adding a few more of them.

diff -r e8367f5cd43f -r be33891b8457 source/encoder/encoder.cpp
--- a/source/encoder/encoder.cpp	Wed Jan 28 11:15:45 2015 -0600
+++ b/source/encoder/encoder.cpp	Wed Jan 28 11:51:48 2015 -0600
@@ -214,7 +214,9 @@
                     if (m_param->rc.rateControlMode == X265_RC_CRF)
                         fprintf(m_csvfpt, "RateFactor, ");
                     fprintf(m_csvfpt, "Y PSNR, U PSNR, V PSNR, YUV PSNR, SSIM, SSIM (dB), "
-                                      "Encoding time, Elapsed time, List 0, List 1\n");
+                                      "List 0, List 1");
+                    /* detailed performance statistics */
+                    fprintf(m_csvfpt, ", Wall time, Total CTU time, Avg WPP\n");
                 }
                 else
                     fputs(summaryCSVHeader, m_csvfpt);
@@ -1159,13 +1161,14 @@
             if (m_param->bEnablePsnr)
                 fprintf(m_csvfpt, "%.3lf, %.3lf, %.3lf, %.3lf,", psnrY, psnrU, psnrV, psnr);
             else
-                fprintf(m_csvfpt, " -, -, -, -,");
+                fputs(" -, -, -, -,", m_csvfpt);
             if (m_param->bEnableSsim)
-                fprintf(m_csvfpt, " %.6f, %6.3f,", ssim, x265_ssim2dB(ssim));
+                fprintf(m_csvfpt, " %.6f, %6.3f", ssim, x265_ssim2dB(ssim));
             else
-                fprintf(m_csvfpt, " -, -,");
-            fprintf(m_csvfpt, " %.3lf, %.3lf", curEncoder->m_frameTime, curEncoder->m_elapsedCompressTime);
-            if (!slice->isIntra())
+                fputs(" -, -", m_csvfpt);
+            if (slice->isIntra())
+                fputs(", -, -", m_csvfpt);
+            else
             {
                 int numLists = slice->isInterP() ? 1 : 2;
                 for (int list = 0; list < numLists; list++)
@@ -1179,10 +1182,14 @@
                 }
 
                 if (numLists == 1)
-                    fprintf(m_csvfpt, ", -");
+                    fputs(", -", m_csvfpt);
             }
+            // detailed frame statistics
+            fprintf(m_csvfpt, ", %.3lf, %.3lf", curEncoder->m_frameTime, curEncoder->m_elapsedCompressTime);
+            if (curEncoder->m_totalActiveWorkerCount)
+                fprintf(m_csvfpt, ", %.3lf", (double)curEncoder->m_totalActiveWorkerCount / curEncoder->m_activeWorkerCountSamples);
             else
-                fprintf(m_csvfpt, ", -, -");
+                fputs(", 1", m_csvfpt);
             fprintf(m_csvfpt, "\n");
         }
 
diff -r e8367f5cd43f -r be33891b8457 source/encoder/frameencoder.cpp
--- a/source/encoder/frameencoder.cpp	Wed Jan 28 11:15:45 2015 -0600
+++ b/source/encoder/frameencoder.cpp	Wed Jan 28 11:51:48 2015 -0600
@@ -237,6 +237,9 @@
     int64_t startCompressTime = x265_mdate();
     Slice* slice = m_frame->m_encData->m_slice;
 
+    m_totalActiveWorkerCount = 0;
+    m_activeWorkerCountSamples = 0;
+
     /* Emit access unit delimiter unless this is the first frame and the user is
      * not repeating headers (since AUD is supposed to be the first NAL in the access
      * unit) */
@@ -779,6 +782,10 @@
         // Does all the CU analysis, returns best top level mode decision
         Mode& best = tld.analysis.compressCTU(*ctu, *m_frame, m_cuGeoms[m_ctuGeomMap[cuAddr]], rowCoder);
 
+        // take a sample of the current active worker count
+        ATOMIC_ADD(&m_totalActiveWorkerCount, m_activeWorkerCount);
+        ATOMIC_INC(&m_activeWorkerCountSamples);
+
         /* advance top-level row coder to include the context of this CTU.
          * if SAO is disabled, rowCoder writes the final CTU bitstream */
         rowCoder.encodeCTU(*ctu, m_cuGeoms[m_ctuGeomMap[cuAddr]]);
diff -r e8367f5cd43f -r be33891b8457 source/encoder/frameencoder.h
--- a/source/encoder/frameencoder.h	Wed Jan 28 11:15:45 2015 -0600
+++ b/source/encoder/frameencoder.h	Wed Jan 28 11:51:48 2015 -0600
@@ -147,6 +147,7 @@
     uint64_t                 m_SSDU;
     uint64_t                 m_SSDV;
     double                   m_ssim;
+    uint64_t                 m_accessUnitBits;
     uint32_t                 m_ssimCnt;
     MD5Context               m_state[3];
     uint32_t                 m_crc[3];
@@ -157,8 +158,10 @@
     FrameStats               m_frameStats;          // stats of current frame for multi-pass encodes
     volatile bool            m_bAllRowsStop;
     volatile int             m_vbvResetTriggerRow;
-    volatile int             m_activeWorkerCount;
-    uint64_t                 m_accessUnitBits;
+
+    volatile int             m_activeWorkerCount;        // count of workers current encoding or filtering CTUs
+    volatile int             m_totalActiveWorkerCount;   // sum of m_activeWorkerCount sampled at end of each CTU
+    volatile int             m_activeWorkerCountSamples; // count of times m_activeWorkerCount was sampled (think vbv restarts)
 
     Encoder*                 m_top;
     x265_param*              m_param;